چكيده به لاتين
With production from the reservoirs and the passage of time, the reservoir pressure decreases in such a way that this reduces the influence of natural mechanisms in fluid movement. Using the gas lift method is an effective way to improve the performance of the reservoir and increase its productivity. Due to the limitation of gas resources as well as the necessity of the optimal distribution of gas between the wells in such a way that the maximum amount of production in the oil fields is achieved, the optimization of gas injection in this process is of great importance. In this thesis, the optimization of gas lift has been discussed. At first, by using Mbal, Prosper, and Gap software, the reservoir, wells, and surface equipment were simulated. For the purpose of integration, all sections are presented in Gap software, and forecasting was done for 6 years without optimization. A net present value of $351,087,876.40 was obtained. This amount confirms the necessity of using the optimization method to reach a higher net present value. After carrying out the sensitivity analysis of the tubing inside diameter, the gas injection rate and the separator pressure were identified as influencing variables on the objective function, i.e., the amount of net present value. In order to carry out the optimization process and also speed up the simulation, proxy models have been developed with the aim of replacing these models with the desired simulator. For this purpose, 154 scenarios designed by the central composite design method were used to develop neural network models as proxy models. In this study, four strong neural networks, including Multi-layer perceptron, Radial basis, General regression, and Cascade forward, were investigated. Then, Genetic algorithms, Particle swarm, Ant colony, and a new algorithm named Gray wolf were used for optimization. According to the mentioned steps, the average relative error of artificial neural network models was 2.24, 2.99, 10.68, and 1.75%, respectively. Therefore, the results showed that the Cascade forward model is the most accurate and efficient for predicting the behavior of the model in this study. In the optimization process, the Gray wolf optimization algorithm brought in the highest amount of current net profit with $788,512,038 and has performed better than other algorithms. The results of this research show that the use of optimization methods leads to a significant net present value and cumulative production of oil compared to the values obtained from non-optimization methods. These results clearly confirm the need and necessity of conducting this type of research in oil fields.
